Segmentation and classification of 3-D spots in FISH images

In this paper, we propose an automated method to segment and classify the 3-D spots in fluorescence in-situ hybridization images from ovarian germline nurse cells of Drosophila melanogaster. The spot segmentation consists of a smoothing step followed by top-hat filtering and 3-D region growing. After the spots are segmented, a number of features such as volume, texture, and contrast are measured so as to classify the real spots from the noise with the help of a Bayesian classifier. Experimental results demonstrate the effectiveness of the proposed scheme in terms of segmentation and classification accuracy.